NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
PyPI, #4162 most downloaded on PyPI
Single-Cell Analysis in Python.
Latest stable
1.12.5
Last release 10 days ago
05 Nov 2026
Ships fairly regularly
a new release about every 2 months
Most releases are documented
notes for 50 of the last 60 stable releases
Nothing withdrawn
no release was ever pulled
9 years old
99 releases · first in 2017
Drag across the ruler to compare, or tap a release to open it.
- a new plotting gallery for visualizing-marker-genes F Ramirez
a new plotting gallery for visualizing-marker-genes F Ramirez
tutorials are integrated on ReadTheDocs, pbmc3k and paga-paul15 A Wolf
CZI’s cellxgene directly reads .h5ad files the cellxgene developers
the UCSC Single Cell Browser requires exporting via cellbrowser() M Haeussler
highly_variable_genes() supersedes sc.pp.filter_genes_dispersion , it gives the same results but, by default, expects logarithmized data and doesn’t subset A Wolf
(v1.3.6)=
2018-12-11visualizing-marker-genes {smaller}F Ramirezpbmc3k and paga-paul15 {smaller}A Wolf.h5ad files {smaller}the cellxgene developers~scanpy.external.exporting.cellbrowser {smaller}M Haeussler~scanpy.pp.highly_variable_genes supersedes sc.pp.filter_genes_dispersion, it gives the same results but, by default, expects logarithmized data and doesn’t subset {smaller}A WolfYour coding agent can read these notes before it upgrades. Set up the MCP server →
- uncountable figure improvements #369 F Ramirez
uncountable figure improvements #369 F Ramirez
(v1.3.5)=
2018-12-09369 {smaller}F Ramirez- leiden() wraps the recent graph clustering package by Traag et al. [ 2019 ] K Polanski
leiden() wraps the recent graph clustering package by Traag et al. [ 2019 ] K Polanski
bbknn() wraps the recent batch correction package [ Polański et al. , 2019 ] K Polanski
calculate_qc_metrics() caculates a number of quality control metrics, similar to calculateQCMetrics from Scater [ McCarthy et al. , 2017 ] I Virshup
(v1.3.4)=
2018-11-24~scanpy.tl.leiden wraps the recent graph clustering package by {cite:t}Traag2019 {smaller}K Polanski~scanpy.external.pp.bbknn wraps the recent batch correction package {cite:p}Polanski2019 {smaller}K Polanski~scanpy.pp.calculate_qc_metrics caculates a number of quality control metrics, similar to calculateQCMetrics from Scater {cite:p}McCarthy2017 {smaller}I Virshup- a fully distributed preprocessing backend T White and the Laserson Lab
a fully distributed preprocessing backend T White and the Laserson Lab
read_10x_h5() and read_10x_mtx() read Cell Ranger 3.0 outputs #334 Q Gong
Note
Also see changes in anndata 0.6.
changed default compression to None in write_h5ad() to speed up read and write, disk space use is usually less critical
performance gains in write_h5ad() due to better handling of strings and categories S Rybakov
(v1.3.3)=
2018-11-05T White and the Laserson Lab~scanpy.io.read_10x_h5 and {func}~scanpy.io.read_10x_mtx read Cell Ranger 3.0 outputs {pr}334 {smaller}Q Gong#### Also see changes in anndata 0.6.
- changed default compression to `None` in {meth}`~anndata.AnnData.write_h5ad` to speed up read and write, disk space use is usually less critical
- performance gains in {meth}`~anndata.AnnData.write_h5ad` due to better handling of strings and categories {smaller}`S Rybakov`
Nothing published for this version
- Scanpy and AnnData support loom’s layers so that computations for single-cell RNA velocity [ La Manno et al. , 2018 ] become feasible S Rybakov and
Scanpy and AnnData support loom’s layers so that computations for single-cell RNA velocity [ La Manno et al. , 2018 ] become feasible S Rybakov and V Bergen
scvelo harmonizes with Scanpy and is able to process loom files with splicing information produced by Velocyto [ La Manno et al. , 2018 ] , it runs a lot faster than the count matrix analysis of Velocyto and provides several conceptual developments
dotplot() for visualizing genes across conditions and clusters, see here #199 F Ramirez
heatmap() for pretty heatmaps #175 F Ramirez
violin() produces very compact overview figures with many panels #175 F Ramirez
magic() for imputation using data diffusion [ van Dijk et al. , 2018 ] #187 S Gigante
sc.external.pp.dca for imputation and latent space construction using an autoencoder [ Eraslan et al. , 2019 ] #186 G Eraslan
(v1.3.1)=
2018-09-03LaManno2018LaManno2018 become feasible {smaller}S Rybakov and V BergenLaManno2018, it runs a lot faster than the count matrix analysis of Velocyto and provides several conceptual developmentspl-generic)~scanpy.pl.dotplot for visualizing genes across conditions and clusters, see here {pr}199 {smaller}F Ramirez~scanpy.pl.heatmap for pretty heatmaps {pr}175 {smaller}F Ramirez~scanpy.pl.violin produces very compact overview figures with many panels {pr}175 {smaller}F Ramirezexternal <../external/index>:~scanpy.external.pp.magic for imputation using data diffusion {cite:p}vanDijk2018 {pr}187 {smaller}S Gigantesc.external.pp.dca for imputation and latent space construction using an autoencoder {cite:p}Eraslan2019 {pr}186 {smaller}G EraslanNothing published for this version
Nothing published for this version
- highest_expr_genes() for quality control; plot genes with highest mean fraction of cells, similar to plotQC of Scater [ McCarthy et al. , 2017 ] #16
highest_expr_genes() for quality control; plot genes with highest mean fraction of cells, similar to plotQC of Scater [ McCarthy et al. , 2017 ] #169 F Ramirez
(v1.2.1)=
2018-06-08pl-generic marker genes and quality control.~scanpy.pl.highest_expr_genes for quality control; plot genes with highest mean fraction of cells, similar to plotQC of Scater {cite:p}McCarthy2017 {pr}169 {smaller}F Ramirez- paga() improved, see PAGA ; the default model changed, restore the previous default model by passing model='v1.0'
paga() improved, see PAGA ; the default model changed, restore the previous default model by passing model='v1.0'
(v1.2.0)=
2018-06-08~scanpy.tl.paga improved, see PAGA; the default model changed, restore the previous default model by passing model='v1.0'- set_figure_params() by default passes vector_friendly=True and allows you to produce reasonablly sized pdfs by rasterizing large scatter plots A Wol
set_figure_params() by default passes vector_friendly=True and allows you to produce reasonablly sized pdfs by rasterizing large scatter plots A Wolf
draw_graph() defaults to the ForceAtlas2 layout [ Chippada, 2018 , Jacomy et al. , 2014 ] , which is often more visually appealing and whose computation is much faster S Wollock
scatter() also plots along variables axis MD Luecken
pca() and log1p() support chunk processing S Rybakov
regress_out() is back to multiprocessing F Ramirez
read() reads compressed text files G Eraslan
mitochondrial_genes() for querying mito genes FG Brundu
mnn_correct() for batch correction [ Haghverdi et al. , 2018 , Kang, 2018 ]
phate() for low-dimensional embedding [ Moon et al. , 2019 ] S Gigante
sandbag() , cyclone() for scoring genes [ Fechtner, 2018 , Scialdone et al. , 2015 ]
(v1.1.0)=
2018-06-01~scanpy.set_figure_params by default passes vector_friendly=True and allows you to produce reasonablly sized pdfs by rasterizing large scatter plots {smaller}A Wolf~scanpy.tl.draw_graph defaults to the ForceAtlas2 layout {cite:p}Jacomy2014,Chippada2018, which is often more visually appealing and whose computation is much faster {smaller}S Wollock~scanpy.pl.scatter also plots along variables axis {smaller}MD Luecken~scanpy.pp.pca and {func}~scanpy.pp.log1p support chunk processing {smaller}S Rybakov~scanpy.pp.regress_out is back to multiprocessing {smaller}F Ramirez~scanpy.io.read reads compressed text files {smaller}G Eraslan~scanpy.queries.mitochondrial_genes for querying mito genes {smaller}FG Brundu~scanpy.external.pp.mnn_correct for batch correction {cite:p}Haghverdi2018,Kang2018~scanpy.external.tl.phate for low-dimensional embedding {cite:p}Moon2019 {smaller}S Gigante~scanpy.external.tl.sandbag, {func}~scanpy.external.tl.cyclone for scoring genes {cite:p}Scialdone2015,Fechtner2018Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
- Scanpy is much faster and more memory efficient: preprocess, cluster and visualize 1.3M cells in 6h , 130K cells in 14min , and 68K cells in 3min A
Scanpy is much faster and more memory efficient: preprocess, cluster and visualize 1.3M cells in 6h , 130K cells in 14min , and 68K cells in 3min A Wolf
the API gained a preprocessing function neighbors() and a class Neighbors() to which all basic graph computations are delegated A Wolf
Warning
Upgrading to 1.0 isn’t fully backwards compatible in the following changes
the graph-based tools louvain() dpt() draw_graph() umap() diffmap() paga() require prior computation of the graph: sc.pp.neighbors(adata, n_neighbors=5); sc.tl.louvain(adata) instead of previously sc.tl.louvain(adata, n_neighbors=5)
install numba via conda install numba , which replaces cython
the default connectivity measure (dpt will look different using default settings) changed. setting method='gauss' in sc.pp.neighbors uses gauss kernel connectivities and reproduces the previous behavior, see, for instance in the example paul15 .
namings of returned annotation have changed for less bloated AnnData objects, which means that some of the unstructured annotation of old AnnData files is not recognized anymore
replace occurances of group_by with groupby (consistency with pandas )
it is worth checking out the notebook examples to see changes, e.g. the seurat example.
upgrading scikit-learn from 0.18 to 0.19 changed the implementation of PCA, some results might therefore look slightly different
UMAP [ McInnes et al. , 2018 ] can serve as a first visualization of the data just as tSNE, in contrast to tSNE, UMAP directly embeds the single-cell graph and is faster; UMAP is also used for measuring connectivities and computing neighbors, see neighbors() A Wolf
graph abstraction: AGA is renamed to PAGA : paga() ; now, it only measures connectivities between partitions of the single-cell graph, pseudotime and clustering need to be computed separately via louvain() and dpt() , the connectivity measure has been improved A Wolf
logistic regression for finding marker genes rank_genes_groups() with parameter method='logreg' A Wolf
louvain() provides a better implementation for reclustering via restrict_to A Wolf
scanpy no longer modifies rcParams upon import, call :func: scanpy.set_figure_params to set the ‘scanpy style’ A Wolf
default cache directory is ./cache/ , set settings.cachedir to change this; nested directories in this are avoided A Wolf
show edges in scatter plots based on graph visualization draw_graph() and umap() by passing edges=True A Wolf
downsample_counts() for downsampling counts MD Luecken
default 'louvain_groups' are called 'louvain' A Wolf
'X_diffmap' contains the zero component, plotting remains unchanged A Wolf
(v1.0.0)= ### 1.0.0 {small}`2018-03-30`
#### Major updates
Scanpy is much faster and more memory efficient: preprocess, cluster and visualize 1.3M cells in [6h], 130K cells in [14min], and 68K cells in [3min] {smaller}`A Wolf`
the API gained a preprocessing function {func}`~scanpy.pp.neighbors` and a class {func}`~scanpy.Neighbors` to which all basic graph computations are delegated {smaller}`A Wolf`
```{warning} #### Upgrading to 1.0 isn’t fully backwards compatible in the following changes
the graph-based tools {func}`~scanpy.tl.louvain` {func}`~scanpy.tl.dpt` {func}`~scanpy.tl.draw_graph` {func}`~scanpy.tl.umap` {func}`~scanpy.tl.diffmap` {func}`~scanpy.tl.paga` require prior computation of the graph: sc.pp.neighbors(adata, n_neighbors=5); sc.tl.louvain(adata) instead of previously sc.tl.louvain(adata, n_neighbors=5)
install numba via conda install numba, which replaces cython
the default connectivity measure (dpt will look different using default settings) changed. setting method='gauss' in sc.pp.neighbors uses gauss kernel connectivities and reproduces the previous behavior, see, for instance in the example [paul15].
namings of returned annotation have changed for less bloated AnnData objects, which means that some of the unstructured annotation of old AnnData files is not recognized anymore
replace occurances of group_by with groupby (consistency with pandas)
it is worth checking out the notebook examples to see changes, e.g. the [seurat] example.
upgrading scikit-learn from 0.18 to 0.19 changed the implementation of PCA, some results might therefore look slightly different
```
#### Further updates
UMAP {cite:p}`McInnes2018` can serve as a first visualization of the data just as tSNE, in contrast to tSNE, UMAP directly embeds the single-cell graph and is faster; UMAP is also used for measuring connectivities and computing neighbors, see {func}`~scanpy.pp.neighbors` {smaller}`A Wolf`
graph abstraction: AGA is renamed to [PAGA](https://github.com/theislab/paga): {func}`~scanpy.tl.paga`; now, it only measures connectivities between partitions of the single-cell graph, pseudotime and clustering need to be computed separately via {func}`~scanpy.tl.louvain` and {func}`~scanpy.tl.dpt`, the connectivity measure has been improved {smaller}`A Wolf`
logistic regression for finding marker genes {func}`~scanpy.tl.rank_genes_groups` with parameter method='logreg' {smaller}`A Wolf`
{func}`~scanpy.tl.louvain` provides a better implementation for reclustering via restrict_to {smaller}`A Wolf`
scanpy no longer modifies rcParams upon import, call scanpy.set_figure_params to set the 'scanpy style' {smaller}`A Wolf`
default cache directory is ./cache/, set settings.cachedir to change this; nested directories in this are avoided {smaller}`A Wolf`
show edges in scatter plots based on graph visualization {func}`~scanpy.tl.draw_graph` and {func}`~scanpy.tl.umap` by passing edges=True {smaller}`A Wolf`
{func}`~scanpy.pp.downsample_counts` for downsampling counts {smaller}`MD Luecken`
default 'louvain_groups' are called 'louvain' {smaller}`A Wolf`
'X_diffmap' contains the zero component, plotting remains unchanged {smaller}`A Wolf`
[14min]: https://github.com/scverse/scanpy_usage/blob/master/170522_visualizing_one_million_cells/logfile_130K.txt [3min]: https://nbviewer.jupyter.org/github/scverse/scanpy_usage/blob/master/170503_zheng17/zheng17.ipynb [6h]: https://github.com/scverse/scanpy_usage/blob/master/170522_visualizing_one_million_cells/ [paul15]: https://nbviewer.jupyter.org/github/scverse/scanpy_usage/blob/master/170502_paul15/paul15.ipynb [seurat]: https://nbviewer.jupyter.org/github/scverse/scanpy_usage/blob/master/170505_seurat/seurat.ipynb
- embed cells using umap() [ McInnes et al. , 2018 ] #92 G Eraslan
embed cells using umap() [ McInnes et al. , 2018 ] #92 G Eraslan
score sets of genes, e.g. for cell cycle, using score_genes() [ Satija et al. , 2015 ] : notebook
(v0.4.4)=
2018-02-26~scanpy.tl.umap {cite:p}McInnes2018 {pr}92 {smaller}G Eraslan~scanpy.tl.score_genes {cite:p}Satija2015:
notebook- clustermap() : heatmap from hierarchical clustering, based on seaborn.clustermap() [ Waskom et al. , 2016 ] A Wolf
clustermap() : heatmap from hierarchical clustering, based on seaborn.clustermap() [ Waskom et al. , 2016 ] A Wolf
only return matplotlib.axes.Axes in plotting functions of sc.pl when show=False , otherwise None A Wolf
(v0.4.3)=
2018-02-09~scanpy.pl.clustermap: heatmap from hierarchical clustering,
based on {func}seaborn.clustermap {cite:p}Waskom2016 {smaller}A Wolfmatplotlib.axes.Axes in plotting functions of sc.pl
when show=False, otherwise None {smaller}A Wolf- amendments in PAGA and its plotting functions A Wolf
Nothing published for this version
- export to SPRING [ Weinreb et al. , 2017 ] for interactive visualization of data: spring tutorial S Wollock
export to SPRING [ Weinreb et al. , 2017 ] for interactive visualization of data: spring tutorial S Wollock
(v0.4.0)=
2017-12-23Weinreb2017 for interactive visualization of data:
spring tutorial {smaller}S Wollock- finding marker genes via rank_genes_groups_violin() improved, see #51 F Ramirez
finding marker genes via rank_genes_groups_violin() improved, see #51 F Ramirez
(v0.3.2)=
2017-11-29~scanpy.pl.rank_genes_groups_violin improved,
see {issue}51 {smaller}F RamirezNothing published for this version
- AnnData gains method AnnData.concatenate A Wolf
AnnData gains method AnnData.concatenate A Wolf
AnnData is available as the separate anndata package P Angerer, A Wolf
results of PAGA simplified A Wolf
Nothing published for this version
- paga() computes an abstracted, coarse-grained (PAGA) graph of the neighborhood graph A Wolf
paga() computes an abstracted, coarse-grained (PAGA) graph of the neighborhood graph A Wolf
paga_compare() plot this graph next an embedding A Wolf
paga_path() plots a heatmap through a node sequence in the PAGA graph A Wolf
(v0.2.9)=
2017-10-25~scanpy.tl.paga computes an abstracted, coarse-grained (PAGA) graph of the neighborhood graph {smaller}A Wolf~scanpy.pl.paga_compare plot this graph next an embedding {smaller}A Wolf~scanpy.pl.paga_path plots a heatmap through a node sequence in the PAGA graph {smaller}A WolfNothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Scanpy includes preprocessing, visualization, clustering, pseudotime and trajectory inference, differential expression testing and simulation of gene
Scanpy includes preprocessing, visualization, clustering, pseudotime and trajectory inference, differential expression testing and simulation of gene regulatory networks. The implementation efficiently deals with datasets of more than one million cells . A Wolf, P Angerer
(v0.2.1)=
2017-07-24Scanpy includes preprocessing, visualization, clustering, pseudotime and
trajectory inference, differential expression testing and simulation of gene
regulatory networks. The implementation efficiently deals with datasets of more
than one million cells. {smaller}A Wolf, P Angerer
One column per quarter.